CN106650020A - Analysis method of complex receptor model pollution source - Google Patents
Analysis method of complex receptor model pollution source Download PDFInfo
- Publication number
- CN106650020A CN106650020A CN201611049694.3A CN201611049694A CN106650020A CN 106650020 A CN106650020 A CN 106650020A CN 201611049694 A CN201611049694 A CN 201611049694A CN 106650020 A CN106650020 A CN 106650020A
- Authority
- CN
- China
- Prior art keywords
- pollutant
- source
- model
- coreceptor
- contribution rate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
Abstract
The invention provides an analysis method of complex receptor model pollution source. The method comprises the steps of pretreating receptor sample data, extracting a principal component factor, decomposing non-negative constraint matrix, and solving the contribution rate of the pollution source. The invention makes the composite use of two receptor models, and uses the non-negative constraint matrix decomposition to backstepping linear independent fingerprint of the pollution source, thus satisfying the requirement of linear independence of the pollution source composition spectrum imposed by the chemical mass balance model (CMB). The method further uses CMB model to solve the contribution rate of the pollution source, therefore the advantages of two models are fully exerted to emphatically enhance the accuracy and reliability of source apportionment results. The analysis method of complex receptor model pollution source can provide necessary technological support in the fields of the environmental quality assessment, environmental risk assessment, water source protection, total emission reduction, ecological restoration, pollution accident investigation and damage compensation, effectively control environmental pollution, and safeguard ecological safety.
Description
Technical field
The invention belongs to environmental protection technical field, and in particular to a kind of coreceptor model pollutes Source Apportionment.
Background technology
In field of environment protection, when Environmental Quality Evalution, environmental risk assessment, pollutant fluxes are carried out, need to pass through
Quantitative approach filters out characteristic contamination and parses pollution sources and its contribution rate, so as to efficiently and effectively control pollution, ensures
Environmental security.So-called pollution source resolution, is to study the serial of methods that pollution sources affect on ambient contamination and act on, and is commonly used
Analytic technique have inventory analysis method, Diffusion Analysis method and receptor model analytic approach.Wherein, inventory analysis method is by seeing
Survey and the source emission amount of simulating pollution thing, discharge characteristics and discharge geographical distribution etc., setting up list model carries out pollution source resolution；
Diffusion Analysis are owned by France in prediction type model, it be the emissions data by being input into each pollution sources and relevant parameter information come
The change in time and space situation of prediction pollutant；Receptor model analyzes rule by the chemistry to acceptor sample and microscopic analysis, it is determined that
One class technology of each pollution sources contribution rate, its final purpose is recognized to the contributive pollution sources of acceptor, and is quantitatively calculated each
The share rate of pollution sources.
Fig. 1 is Diffusion Analysis process schematic in prior art.As shown in figure 1, being polluted using diffusion model
Source resolution, not only needs complete emission inventory, it will also be appreciated that the blowdown feature of emission source, emission index, drain time, acceptor
Discharge parameter in position etc..As can be seen that carrying out that source resolution is relatively costly using diffusion model, difficulty is larger.
Fig. 2 is receptor model analysis process schematic in prior art.As shown in Fig. 2 the advantage of receptor model is it
Using real data receptor come driving model, quantitative source resolution can be carried out to each sample, it is also possible to according to fitness
Analysis carrys out the error in identification model source and real source.Its inferior position is that model will have acceptor sample concentration, in addition also require to have with
The corresponding uncertain value of acceptor density, it is desirable to source line independence, it is impossible to which effective district is exceptionally originated and local source.Receptor model is opened
Exhibition pollution source resolution, needs to meet several basic assumption conditions：(1) source line is fixed；(2) data receptor acquisition process and condition
Do not change significantly；(3) data receptor is representative；(4) pollution sources existence time or spatial variability.With diffusion model phase
Than, the input data source of receptor model is relatively easy to, it is not necessary to which complete source spectrum inventory, model operating cost is relatively low, meanwhile,
As effective supplement of diffusion model, people can be made by the cognition and comprehension source to deviation and the relation of acceptor.
Comparatively, receptor model does not rely on the data such as emission source conditions of discharge and hydrology-water quality, without following the trail of pollution
The transition process of thing, and compared with the diffusion model of prediction type, a kind of receptor model diagnosis formula model at last, it explain in the past and
It is not future, compares and be adapted to the source resolution for realizing pollution sources.According to whether grasping detailed discharge of pollutant sources in advance
Characteristic spectrum, receptor model can be divided into chemotherapy synergism (CMB), multivariate statistical model, UNMIX models etc..With environment
Monitoring network it is increasingly perfect, monitoring means, the continuous progress of methods and techniques, researcher can obtain and in a large number and accurately see
Survey sample data, this provides guarantee to carry out the quantitative study for polluting source resolution, receptor model Source apportionment obtained compared with
Fast development and application, are widely used for carrying out the dirt of various pollutants in the surrounding mediums such as air, water body, deposit and soil
Dye source resolution research.
Wherein, chemotherapy synergism is the more ripe source resolution model of class development, and it does not rely on emission source
The data such as condition, meteorology, landform, without the transition process for following the trail of particulate matter, modular concept understands, it is easy to receive, and is using most
For a kind of extensive receptor model.It is more successful for the source resolution of inorganic pollution, and by EPA nothing is set to
The prefered method of machine thing source resolution.The model sets up pollution sources analytical quantitative models according to mass conservation law, by measuring source
Physics, chemical property with acceptor sample, qualitative recognition is to the contributive pollution sources of acceptor and quantitatively calculates each pollution sources and divides
Load rate, its theoretical foundation is that the dactylogram of each pollution sources has certain difference, such that it is able to pass through to detect each in acceptor
The content (composition) of kind of material is determining the contribution rate of each pollution sources.But chemotherapy synergism application needs to meet multinomial vacation
If：(1) all pollution sources classes to there is obvious contribution in environment acceptor, and the pollutant that each source class is discharged can be identified
Chemical composition has obvious difference；(2) the pollutant chemistry composition that each source class is discharged is relatively stable, and what chemical group was divided asks without bright
Development rings；(3) without interacting between the pollutant that each source class is discharged, the change in transmitting procedure can be ignored；
(4) all pollution sources component spectrums are linear independences；(5) pollute source category and be less equal than chemical constituent species；(6) measure
Uncertainty is random, meets normal distribution.What is more important, the acquisition difficulty of pollution sources dactylogram is larger, cost compared with
Height, needs being continually changing according to environment, follows the trail of emission inventory, discharge characteristics fingerprint base is updated, particularly the currently monitored
Network is not that under the reality for covering all disposal of pollutants points, condition known to pollution sources dactylogram is often difficult to meet.
From unlike chemotherapy synergism, polynary receptor model does not need previously known specific pollution derived components
Spectrum, but pollution sources dactylogram is extracted by Factorization, and obtain the source contribution rate of each pollution factor.Polynary receptor model
Data acquisition system can be parsed, compressed data dimension analyzes the relation between multiple variables, and this characteristic is for big discharge observation data
Analysis it is extremely useful, it is possible to use the less representational factor is illustrating the main information of numerous variables.However, polynary
Acceptor source resolution model there is also some defects, and for example, Factorization can not obtain the factor loading of " unique " and factor score,
The factor loading extracted by factorial analysis and factor score can rotate " the uncomfortable fixed solution " that obtain various ways by orthogonal,
Factor loading that prior utilization factorial analysis is obtained and usually there is negative value in factor score, this not phase with practical application
Symbol, because the derived components spectrum or source contribution rate of negative value are without actual physical significance；Positive definite matrix decomposition model although it is contemplated that
The uncertain problem of data receptor, but the pollution sources contribution rate for ensureing to extract by introducing penalty factor and confactor
The uncertain problem of discharge of pollutant sources is not accounted for.
The content of the invention
The technical problem to be solved in the present invention is to provide a kind of coreceptor model pollution Source Apportionment, existing to solve
There is the deficiency that Source Apportionment is polluted in technology, improve the accuracy and reliability of source resolution result.
According to an aspect of the invention, there is provided a kind of coreceptor model pollutes Source Apportionment, methods described bag
Include：
Step S1, the pretreatment of acceptor sample data；
Step S2, extracts principal component factor number；
Step S3, decomposes nonnegativity restrictions matrix；
Step S4, solves pollution sources contribution rate.
In such scheme, step S1 is further included：
Step S101, from the angle of data processing, audits acceptor sample data；
Step S102, from the angle of pollutant, selects coreceptor model variable；
Step S103, is standardized to the model variable.
In such scheme, the angle Selection coreceptor model variable from pollutant specifically includes following process：
First, it is impossible to while selecting activity strong, or there is the pollutant of reaction；
Second, contaminated accident impact and the pollutant that detects cannot be used for model calculating；
3rd, it is impossible to while selecting to contain the pollutant that mutual conversion can occur in identical element, and transition process；
4th, concentration do not detect or shortage of data pollutant, suitable uncertainty value is replaced and arranged by data；
5th, pollutant has concentration value but the quality of data is unknown, by being analyzed to determine whether to investigate the pollutant again.
In such scheme, principal component ratio characteristics are in step S2：
Characteristic value is more than 1；
Cumulative proportion in ANOVA value is more than 85%；
The coefficient of determination is more than 0.9；
Exter functional values are less than 0.1.
In such scheme, step S2 is further included：
Step S201, solves covariance matrix Z：Z=DDt, D is data receptor collection；
Step S202, solves characteristic value E and characteristic vector Q：Z=QEQt；
Step S203, solves Factor load-matrix S without spin：S=QE1/2；
Step S204, solves factor score matrix R without spin：R=(StS)-1StD；
Step S205, generates characteristic value, cumulative variance, the coefficient of determination and Exner function criterion matrixes, extract principal component because
Subnumber.
In such scheme, nonnegativity restrictions matrix is decomposed in step S3, decomposed by the improved alternately Return Law.
In such scheme, the improved alternately Return Law is further included：
Step S301, k=0 arranges initial value S (the k)=S of two unknown matrixes0With R (k)=R0；
Step S302, keeps R=R (k) constant, seeks suitable Δ S so that S=S (k)+Δ S minimizes Q value (D-
SR)；
Step S303, keeps S=S (k)+Δ S constant, seeks suitable Δ R so that R=R (k)+Δ R minimizes Q values；
Step S304, seeks suitable spreading coefficient α so that R=R (k)+α Δ R, and S=S (k)+α Δs S minimizes Q
Value；
Step S305, k=k+1, repeat step (b)～(c) are until iteration convergence.
In such scheme, step S4 is further included：
The standard deviation for assuming first that pollutant in certain pollution sources is linearly dependent in the pollution sources tribute for specifying pollutant
Ratio is offered, that is, is had：
Wherein：σiThe standard deviation of the i-th pollutant in the derived components spectrum that expression is extracted；δiIt is the i-th class in acceptor sample
The standard deviation of pollutant；The concentration of emission of the i-th pollutant in the derived components spectrum that expression is extracted；DiRepresent acceptor sample
In the i-th pollutant concentration；
Secondly, source contribution rate process is calculated as follows：
Step S401, it is assumed that initial source contribution rateWherein n represents pollution sources number；
Step S402, calculates effective variance diagonal matrixIts element is：
Wherein：For the uncertain standard deviation of i-th kind of pollutant；For i-th kind of pollutant of jth class pollution sources
The uncertain standard deviation of concentration；For the source contribution rate that last time tries to achieve；
Step S403, source contribution rate is calculated：
Wherein：S is pollution sources finger-print；ST is its transposed matrix；The effective variance obtained for step (b) is diagonal
Battle array；D is sample concentration；
Step S404, source contribution rate iteration deviation is calculated：
If deviation is more than certain setting error precision (such as 0.001), repeat step S401～403；Otherwise, terminate changing
Generation.
In such scheme, the calculating source contribution rate can also include：
Step S405, passes through
Calculate source contribution rate uncertainty deviation.
Two receptoroid models are carried out Application of composite by pollution Source Apportionment of the present embodiment based on coreceptor model,
Propose a kind of pollution Source Apportionment based on nonnegativity restrictions matrix decomposition Chemical mass balance mode composite model.The present invention is utilized
The anti-pollution sources finger-print for releasing linear independence of nonnegativity restrictions matrix decomposition, to meet chemotherapy synergism to pollution sources
The Line independent of component spectrum is required, and calculates pollution sources contribution rate using CMB models, so as to give full play to two kinds of models each
Advantage, put forth effort improve source resolution result accuracy and reliability.The present invention can be commented for Environmental Quality Evalution, environmental risk
Valency, water resource protection, total amount reduction of discharging, restoration of the ecosystem, contamination accident investigation, damages provide necessary science and technology support, for having
The control environmental pollution of effect ground, ensures that ecological safety has important practical significance.The Source Apportionment coupling that achievement of the present invention is proposed
The thinking of application is closed, also there is certain scientific value to pollution sources analytic theory and the perfect of method.
Description of the drawings
Fig. 1 is diffusion model pollution sources resolving schematic diagram in prior art；
Fig. 2 is receptor model pollution sources resolving schematic diagram in prior art；
Fig. 3 is embodiment of the present invention coreceptor model pollution Source Apportionment process schematic；
Fig. 4 is embodiment of the present invention nonnegativity restrictions matrix decomposition Chemical mass balance mode coreceptor model algorithm logic chart.
Specific embodiment
To make the technical problem to be solved in the present invention, technical scheme and advantage clearer, below in conjunction with accompanying drawing and tool
Body embodiment is described in detail.
By the analysis to prior art, pollution source resolution is carried out based on single receptor model analytic technique, deposited
Many not enough.For example, chemotherapy synergism is more accurately to pollute source discrimination using comparative maturity, parsing,
But its application conditions is stricter, in addition to needing to be known a priori by pollution sources finger-print, can't occur between each pollution sources
" mixed source " phenomenon, this is often inconsistent with practical application condition.And polynary receptor model counter can release pollution sources fingerprint image
Spectrum, tries to achieve factor score, but is based entirely on " ill-posedness " that the derived components spectrum and source contribution rate of factorial analysis often have solution
Problem, needs by increasing additional conditions (such as nonnegativity restrictions, penalty, R items) to compress solution space, nonetheless,
Can not simultaneously consider the uncertain problem of acceptor density and discharge of pollutant sources concentration using Factorization, weaken source contribution rate
The precision and reliability of calculating.
For the limitation of the pollution Source Apportionment of single receptor model, the present invention proposes a kind of coreceptor model
Pollution Source Apportionment, i.e. nonnegativity restrictions matrix decomposition Chemical mass balance mode composite model (Nonnegative Matrix
Factorization Chemical Mass Balance, referred to as NMFCMB).The model utilizes nonnegativity restrictions matrix decomposition
The discharge of pollutant sources component spectrum and corresponding uncertainty value of linear independence are obtained, to meet chemotherapy synergism to pollution
The Line independent of derived components spectrum requires that recycling chemotherapy synergism calculates corresponding pollution sources contribution rate, so as to fill
The respective advantage of CMB models and Factor Analysis Model is waved in distribution, and the optimization for reaching pollution source resolution is processed.
Fig. 3 pollutes the process schematic of Source Apportionment for the coreceptor model of the specific embodiment of the invention.Such as Fig. 3 institutes
Show that the coreceptor model pollution Source Apportionment of the present embodiment comprises the steps：
Step S1, the pretreatment of acceptor sample data；
Step S2, extracts principal component factor number；
Step S3, decomposes nonnegativity restrictions matrix；
Step S4, solves pollution sources contribution rate.
Fig. 4 is the present embodiment nonnegativity restrictions matrix decomposition Chemical mass balance mode coreceptor model algorithm logic chart.Such as Fig. 4
It is shown：
Step S1 is specifically included：
Step S101, from the angle of data processing, audits acceptor sample data.
Firstly the need of explanation, acceptor sample data here, the concentration measurement data of pollutant sample is referred mainly to, wrapped
Include and do not detect item, disappearance item, concentration measurement etc..Therefore, the present embodiment can also include step S0, collect concentration of specimens number
According to.Here concentration of specimens data can be expressed as D=m × n, and D is sample data matrix.
Analysis of uncertainty includes the uncertainty of model itself and the uncertainty of mode input.It is compound in the present invention
Receptor model, using enough acceptor sample sets, using Multivariate statistical techniques, extracts pollution sources component spectrum, calculates source contribution.Very
Substantially, the precision of the model calculation is had great significance as the acceptor sample data of mode input item.Using in,
Acceptor sample data set can be the pollutant levels matrix of time and/or spatial sequence.Regardless of whether being time or space sequence
The pollutant levels collection of row, its quality of data can always be sampled, be measured, lab analysis, data record (computer typing),
The impact of the processes such as data copy.For example, the sample of different sampling stages collection, may be by environmental factor, discharge periodic characteristic
Affect larger variation occur.In addition, there is larger regional differentiation, automatic monitoring capability in current monitoring network (ability)
Also it is not very strong, most routine monitoring is still carried out by manual type.Therefore, measuring instrument, analysis method, data note
Record, data copy also may all bring the uncertain error of acceptor density value.Further, the pollutant in acceptor sample comes from
The contribution of different pollution sources, and the pollutant of discharge of pollutant sources there is also certain uncertainty, such as thing from source to acceptor process
Reason, chemical, biodegradable etc..Used as the acceptor sample of receptor model input item, there are three kinds of situations in its numerical value：Measurement concentration
Value, (less than detection limit) and null value (loss of data or without measure) are not detected.
Therefore, data examination ＆ verification here, including following three aspect：
When there is concentration measurement, there is experience to know pollutant levels measured value according to environmental survey and
Not, anomaly is judged whether, its uncertainty value considers analysis uncertainty and method detection limit；The data are examined
Core, is processed not detecting item；To the process without measured value item.
For item is not detected, substituted with the half of detection limit, and corresponding uncertain then value is the 5/6 of detection limit
Times；
For null value, processing mode can be realized by following three kinds of modes：First, reject null value sample；Second, comprising
The pollutant sample data variable of null value does not include model calculating；3rd, with pollutant levels measured value arithmetic average or geometry
It is average to replace null value, and give the uncertain value of mean concentration three-to-four-fold.
Step S102, from the angle of pollutant, selects coreceptor model variable.
When carrying out pollution source resolution based on coreceptor model in the present invention, its model variable refers to different pollutants
Concentration value, therefore the concentration value of suitable pollutant is selected as the model variable impact important for the model calculation has.
It is described to select suitable pollutant, including：
First, according to the basic assumption of coreceptor model, process of the pollutant from source to acceptor can not be sent out each other
Raw reaction, therefore can not simultaneously select those activity stronger, or there is the pollutant of reaction；
Second, due to the coreceptor model assumption of the present invention, the acceptor sample of different time collection, its pollution concentration is received
The impact of identical pollution sources, therefore, contaminated accident impact and the pollutant that detects cannot be used for model calculating；
3rd, for containing the mutual pollutant (i.e. contamination index) changed can occur in identical element, and transition process
Can not select simultaneously, such as ammonia nitrogen, the concentration value of total nitrogen can not simultaneously be selected as model variable；
4th, concentration do not detect or shortage of data pollutant, suitable uncertainty value is replaced and arranged by data
Reduce its impact to model result；
5th, pollutant has concentration value (measured value or pretreatment values), but the quality of data is unknown, needs by deeply dividing
Analyse to determine whether to investigate the pollutant.
Step S103, is standardized to model variable.
To solve the dimension mismatch problem of different pollutant levels, the deviation of analysis process is eliminated, needed to selected
Sample data be that model variable is standardized.Data normalization of the present invention can by standardizing average values method or
Standard of index method is carried out, and obtains normal data.Such as：
Step S2, extracts principal component factor number, specifically includes：
The present embodiment principal component factor number can according to characteristic value, cumulative proportion in ANOVA, the coefficient of determination, Exner functions, from
Determined by spending.The principal component factor is determined first.The principal component ratio characteristics that the present embodiment is chosen include：Characteristic value is more than 1
All factors are used as the principal component factor；If cumulative proportion in ANOVA value is more than 85%, it is possible to think the result of factorial analysis
Explain enough variable informations；The coefficient of determination is more than components Factor based on 0.9 can determine.Particularly, for Exner
Functional value, in theory can be from 0 to infinity, and its value is acceptable peak for 0.5, thinks dry straight less than 0.1.
The process for extracting principal component factor number is as follows：
Step S201, solves covariance matrix Z：Z=DDt；
Step S202, solves characteristic value E and characteristic vector Q：Z=QEQt；
Step S203, solves Factor load-matrix S without spin：S=QE1/2；
Step S204, solves factor score matrix R without spin：R=(StS)-1StD；
Step S205, generates characteristic value, cumulative variance, the coefficient of determination and Exner function criterion matrixes, extract principal component because
Subnumber.
Step S3, decomposes nonnegativity restrictions matrix, further includes：
In prior art, the most common algorithm of Factorization is exactly alternately to return least square method, and the algorithm is assumed first that
Wherein some objective matrix is, it is known that seek another unknown matrix so that target function value is minimum；Then another is again assumed that
Objective matrix is, it is known that repeat said process；So repeatedly until tending to convergence.It is obvious that the algorithm is simple, but efficiency is low, meter
Calculation amount is larger.In order to improve the computational efficiency of the alternately Return Law, the present embodiment takes a kind of improved alternately Return Law, described to change
The alternating regression algorithm for entering, by realizing that the synchronization of two unknown matrixes is incremented by, so as to reach the target for lifting convergence efficiency, its
Basic process is as follows：
Step S301, k=0 arranges initial value S (the k)=S of two unknown matrixes0With R (k)=R0；
Step S302, keeps R=R (k) constant, seeks suitable Δ S so that S=S (k)+Δ S minimizes Q values；
Step S303, keeps S=S (k)+Δ S constant, seeks suitable Δ R so that R=R (k)+Δ R minimizes Q values；
Step S304, seeks suitable spreading coefficient α so that R=R (k)+α Δ R, and S=S (k)+α Δs S minimizes Q
Value；
Step S305, k=k+1, repeat step (b)～(c) are until iteration convergence.
Step S4, solves pollution sources contribution rate, further includes：
Step S41, using chemotherapy synergism calculate pollution sources contribution rate need provide acceptor sample concentration and its
Uncertainty, and discharge of pollutant sources component spectrum (i.e. dactylogram) and its uncertainty.In the NMFCMB models of the present embodiment,
The uncertain value calculating method of the pollution sources component spectrum instead released using nonnegativity restrictions matrix decomposition is as follows：It is assumed that certain pollution sources
The standard deviation of middle pollutant is linearly dependent in the pollution sources contribution proportion for specifying pollutant, that is, have：
Wherein：σiThe standard deviation of the i-th pollutant in the derived components spectrum that expression is extracted；δiIt is the i-th class in acceptor sample
The standard deviation of pollutant；The concentration of emission of the i-th pollutant in the derived components spectrum that expression is extracted；DiRepresent acceptor sample
In the i-th pollutant concentration.
Step S42, source contribution rate is calculated and calculates source contribution rate using effective variance least square method, and its step is as follows：
Step S401, it is assumed that initial source contribution rateWherein n represents pollution sources number.
Step S402, calculates effective variance diagonal matrixIts element is：
Wherein：For the uncertain standard deviation of i-th kind of pollutant；For i-th kind of pollutant of jth class pollution sources
The uncertain standard deviation of concentration；For the source contribution rate that last time tries to achieve.
Step S403, source contribution rate is calculated：
Wherein：S is pollution sources finger-print；ST is its transposed matrix；The effective variance obtained for step (b) is diagonal
Battle array；D is sample concentration.
Step S404, source contribution rate iteration deviation is calculated：
If deviation is more than certain setting error precision (such as 0.001), repeat step S401～403；Otherwise, terminate changing
Generation.
Step S405, calculates source contribution rate uncertainty deviation：
The present embodiment can also include step S5, carry out certificate authenticity.
Specifically, the one kind in residual variance, mean variance, mass percent, T statistical values can be passed through in the present embodiment
Or it is various carry out certificate authenticity, so as to obtain final source contribution rate.
As can be seen from the above technical solutions, pollution Source Apportionment of the present embodiment based on coreceptor model, by two
Receptoroid model carries out Application of composite, it is proposed that a kind of dirt based on nonnegativity restrictions matrix decomposition Chemical mass balance mode composite model
Dye Source Apportionment.The present invention utilizes the anti-pollution sources finger-print for releasing linear independence of nonnegativity restrictions matrix decomposition, to meet
Chemotherapy synergism is required the Line independent of pollution sources component spectrum, and calculates pollution sources contribution rate using CMB models,
So as to give full play to the respective advantage of two kinds of models, put forth effort the accuracy and reliability for improving source resolution result.The present invention both may be used
Think that Environmental Quality Evalution, environmental risk assessment, water resource protection, total amount reduction of discharging, restoration of the ecosystem, contamination accident investigation, infringement are paid for
The necessary science and technology support of offer is provided, for environmental pollution is efficiently controlled, ensures that ecological safety has important practical significance.This
The thinking of the Source Apportionment coupling application that invention achievement is proposed is perfect also with certain with method to pollution sources analytic theory
Scientific value.
The above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, on the premise of without departing from principle of the present invention, some improvements and modifications can also be made, these improvements and modifications
Should be regarded as protection scope of the present invention.
Claims (9)
1. a kind of coreceptor model pollutes Source Apportionment, it is characterised in that methods described includes：
Step S1, the pretreatment of acceptor sample data；
Step S2, extracts principal component factor number；
Step S3, decomposes nonnegativity restrictions matrix；
Step S4, solves pollution sources contribution rate.
2. coreceptor model according to claim 1 pollutes Source Apportionment, it is characterised in that step S1 enters one
Step includes：
Step S101, from the angle of data processing, audits acceptor sample data；
Step S102, from the angle of pollutant, selects coreceptor model variable；
Step S103, is standardized to the model variable.
3. described coreceptor model according to claim 2 pollutes Source Apportionment, it is characterised in that described from dirt
The angle Selection coreceptor model variable of dye thing, specifically includes following process：
First, it is impossible to while selecting activity strong, or there is the pollutant of reaction；
Second, contaminated accident impact and the pollutant that detects cannot be used for model calculating；
3rd, it is impossible to while selecting to contain the pollutant that mutual conversion can occur in identical element, and transition process；
4th, concentration do not detect or shortage of data pollutant, suitable uncertainty value is replaced and arranged by data；
5th, pollutant has concentration value but the quality of data is unknown, by being analyzed to determine whether to investigate the pollutant again.
4. described coreceptor model according to claim 1 pollutes Source Apportionment, it is characterised in that the step
Principal component ratio characteristics are in S2：
Characteristic value is more than 1；
Cumulative proportion in ANOVA value is more than 85%；
The coefficient of determination is more than 0.9；
Exner functions<0.1.
5. described coreceptor model according to claim 4 pollutes Source Apportionment, it is characterised in that the step
S2 is further included：
Step S201, solves covariance matrix Z：Z=DDt, D is data receptor collection；
Step S202, solves characteristic value E and characteristic vector Q：Z=QEQt；
Step S203, solves Factor load-matrix S without spin：S=QE1/2；
Step S204, solves factor score matrix R without spin：R=(StS)-1StD；
Step S205, generates characteristic value, cumulative variance, the coefficient of determination and Exner function criterion matrixes, extracts the principal component factor
Number.
6. described coreceptor model according to claim 1 pollutes Source Apportionment, it is characterised in that the step
Decompose nonnegativity restrictions matrix in S3, decomposed by the improved alternately Return Law.
7. described coreceptor model according to claim 6 pollutes Source Apportionment, it is characterised in that the improvement
The alternating Return Law, further include：
Step S301, k=0 arranges initial value S (the k)=S of two unknown matrixes0With R (k)=R0；
Step S302, keeps R=R (k) constant, seeks suitable Δ S so that S=S (k)+Δ S minimizes Q values；
Step S303, keeps S=S (k)+Δ S constant, seeks suitable Δ R so that R=R (k)+Δ R minimizes Q values；
Step S304, seeks suitable spreading coefficient α so that R=R (k)+α Δ R, and S=S (k)+α Δs S minimizes Q values；
Step S305, k=k+1, repeat step (b)～(c) are until iteration convergence.
8. described coreceptor model according to claim 1 pollutes Source Apportionment, it is characterised in that the step
S4, further includes：
The standard deviation for assuming first that pollutant in certain pollution sources is linearly dependent in the pollution sources contribution ratio for specifying pollutant
Example, that is, have：
Wherein：σiThe standard deviation of the i-th pollutant in the derived components spectrum that expression is extracted；δiIt is the i-th class pollution in acceptor sample
The standard deviation of thing；The concentration of emission of the i-th pollutant in the derived components spectrum that expression is extracted；DiRepresent the in acceptor sample
The concentration of i pollutants；
Secondly, source contribution rate process is calculated as follows：
Step S401, it is assumed that initial source contribution rateJ=1, wherein 2 ... n, n represent pollution sources number；
Step S402, calculates effective variance diagonal matrixIts element is：
Wherein：For the uncertain standard deviation of i-th kind of pollutant；For i-th kind of pollutant levels of jth class pollution sources
Uncertain standard deviation；For the source contribution rate that last time tries to achieve；
Step S403, source contribution rate is calculated：
Wherein：S is pollution sources finger-print；ST is its transposed matrix；For effective variance diagonal matrix that step (b) is obtained；D is
Sample concentration；
Step S404, source contribution rate iteration deviation is calculated：
If deviation is more than certain setting error precision (such as 0.001), repeat step S401～403；Otherwise, iteration is terminated.
9. described coreceptor model according to claim 8 pollutes Source Apportionment, it is characterised in that the calculating
Source contribution rate can also include：
Step S405, passes through
Calculate source contribution rate uncertainty deviation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611049694.3A CN106650020A (en) | 2016-11-24 | 2016-11-24 | Analysis method of complex receptor model pollution source |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611049694.3A CN106650020A (en) | 2016-11-24 | 2016-11-24 | Analysis method of complex receptor model pollution source |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106650020A true CN106650020A (en) | 2017-05-10 |
Family
ID=58811893
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611049694.3A Pending CN106650020A (en) | 2016-11-24 | 2016-11-24 | Analysis method of complex receptor model pollution source |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106650020A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108195727A (en) * | 2017-11-16 | 2018-06-22 | 苏州科技大学 | Airborne fine particulate matter discharge source localization method based on pca model |
| CN108446531A (en) * | 2018-03-08 | 2018-08-24 | 南开大学 | Receptor Si and Al reconstructing method based on actual measurement derived components spectrum and Source apportionment |
| CN109085291A (en) * | 2018-07-30 | 2018-12-25 | 南开大学 | It lacks component iterative inversion and demarcates nesting-PMF source resolution algorithm |
| CN109711674A (en) * | 2018-12-03 | 2019-05-03 | 北京师范大学 | A kind of finger-print base construction method and device traced to the source for lake and reservoir water pollution |
| CN110738589A (en) * | 2019-10-25 | 2020-01-31 | 济南大学 | method for analyzing underground water chlorinated hydrocarbon pollution source |
| CN111798928A (en) * | 2020-06-30 | 2020-10-20 | 中科三清科技有限公司 | Atmospheric particulate pollution source analysis method and device |
| CN112051362A (en) * | 2020-09-01 | 2020-12-08 | 杭州壹丈科技有限公司 | Wind rose pre-judgment pollutant change model |
| CN112102895A (en) * | 2020-05-27 | 2020-12-18 | 北京师范大学 | Road sediment polycyclic aromatic hydrocarbon source analysis method in coupling migration and conversion process |
| CN112735539A (en) * | 2021-01-18 | 2021-04-30 | 中国科学院生态环境研究中心 | Source analysis method of polluted site characteristic pollutants based on composite receptor model |
| CN113657698A (en) * | 2020-05-12 | 2021-11-16 | 中国环境科学研究院 | Basin partition pollution source identification method based on multivariate statistics and receptor model |
| CN114357751A (en) * | 2021-12-27 | 2022-04-15 | 重庆市生态环境大数据应用中心 | Method for accurately identifying type of cross-section scale river-entering pollution source |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103065198A (en) * | 2012-12-17 | 2013-04-24 | 天津市环境保护科学研究院 | Atmosphere fetor pollution fine source apportionment method |
| CN105158353A (en) * | 2015-07-29 | 2015-12-16 | 北华航天工业学院 | Source apportionment method for polycyclic aromatic hydrocarbon pollution in soil |
| CN105184000A (en) * | 2015-09-18 | 2015-12-23 | 北华航天工业学院 | Nonnegative-constrain-factor pollution source apportionment method based on naive Bayesian source identification |
| CN105868479A (en) * | 2016-04-05 | 2016-08-17 | 北华航天工业学院 | Polycyclic aromatic hydrocarbon source apportionment method |
-
2016
- 2016-11-24 CN CN201611049694.3A patent/CN106650020A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103065198A (en) * | 2012-12-17 | 2013-04-24 | 天津市环境保护科学研究院 | Atmosphere fetor pollution fine source apportionment method |
| CN105158353A (en) * | 2015-07-29 | 2015-12-16 | 北华航天工业学院 | Source apportionment method for polycyclic aromatic hydrocarbon pollution in soil |
| CN105184000A (en) * | 2015-09-18 | 2015-12-23 | 北华航天工业学院 | Nonnegative-constrain-factor pollution source apportionment method based on naive Bayesian source identification |
| CN105868479A (en) * | 2016-04-05 | 2016-08-17 | 北华航天工业学院 | Polycyclic aromatic hydrocarbon source apportionment method |
Non-Patent Citations (4)
| Title |
|---|
| 周炎等: "典型污染时段鹤山大气VOCs的抽样生成潜势及来源解析", 《环境监控与预警》 * |
| 李娇等: "拉林河流域土壤重金属污染特征及来源解析", 《农业工程学报》 * |
| 陈海洋等: "基于NMF与CMB耦合应用的水体污染源解析方法", 《环境科学学报》 * |
| 陈锋: "松花江流域水环境中多环烃的源解析方法与应用研究", 《中国博士学位论文全文数据库》 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108195727A (en) * | 2017-11-16 | 2018-06-22 | 苏州科技大学 | Airborne fine particulate matter discharge source localization method based on pca model |
| CN108446531B (en) * | 2018-03-08 | 2021-06-29 | 南开大学 | Receptor Si and Al reconstruction method based on actual measurement source component spectrum and source analysis technology |
| CN108446531A (en) * | 2018-03-08 | 2018-08-24 | 南开大学 | Receptor Si and Al reconstructing method based on actual measurement derived components spectrum and Source apportionment |
| CN109085291A (en) * | 2018-07-30 | 2018-12-25 | 南开大学 | It lacks component iterative inversion and demarcates nesting-PMF source resolution algorithm |
| CN109085291B (en) * | 2018-07-30 | 2020-12-01 | 南开大学 | Missing component iterative inversion calibration nesting-PMF source analysis algorithm |
| CN109711674A (en) * | 2018-12-03 | 2019-05-03 | 北京师范大学 | A kind of finger-print base construction method and device traced to the source for lake and reservoir water pollution |
| CN110738589A (en) * | 2019-10-25 | 2020-01-31 | 济南大学 | method for analyzing underground water chlorinated hydrocarbon pollution source |
| CN113657698A (en) * | 2020-05-12 | 2021-11-16 | 中国环境科学研究院 | Basin partition pollution source identification method based on multivariate statistics and receptor model |
| CN112102895A (en) * | 2020-05-27 | 2020-12-18 | 北京师范大学 | Road sediment polycyclic aromatic hydrocarbon source analysis method in coupling migration and conversion process |
| CN112102895B (en) * | 2020-05-27 | 2024-01-26 | 北京师范大学 | Road sediment polycyclic aromatic hydrocarbon source analysis method for coupling migration conversion process |
| CN111798928B (en) * | 2020-06-30 | 2020-12-29 | 中科三清科技有限公司 | Atmospheric particulate pollution source analysis method and device |
| CN111798928A (en) * | 2020-06-30 | 2020-10-20 | 中科三清科技有限公司 | Atmospheric particulate pollution source analysis method and device |
| CN112051362A (en) * | 2020-09-01 | 2020-12-08 | 杭州壹丈科技有限公司 | Wind rose pre-judgment pollutant change model |
| CN112051362B (en) * | 2020-09-01 | 2022-06-24 | 杭州壹丈科技有限公司 | Wind rose pre-judgment pollutant change model |
| CN112735539A (en) * | 2021-01-18 | 2021-04-30 | 中国科学院生态环境研究中心 | Source analysis method of polluted site characteristic pollutants based on composite receptor model |
| CN112735539B (en) * | 2021-01-18 | 2024-04-16 | 中国科学院生态环境研究中心 | Source analysis method for characteristic pollutants of polluted site based on composite receptor model |
| CN114357751A (en) * | 2021-12-27 | 2022-04-15 | 重庆市生态环境大数据应用中心 | Method for accurately identifying type of cross-section scale river-entering pollution source |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106650020A (en) | Analysis method of complex receptor model pollution source | |
| CN105158353B (en) | Source apportionment method for polycyclic aromatic hydrocarbon pollution in soil | |
| Swenson et al. | Phylogenetic and functional alpha and beta diversity in temperate and tropical tree communities | |
| Nkoa et al. | Weed abundance, distribution, diversity, and community analyses | |
| KR101811270B1 (en) | Method and system for checking goods | |
| CN105631203A (en) | Method for recognizing heavy metal pollution source in soil | |
| CN105868479A (en) | Polycyclic aromatic hydrocarbon source apportionment method | |
| Uber et al. | Comparing alternative tracing measurements and mixing models to fingerprint suspended sediment sources in a mesoscale Mediterranean catchment | |
| CN108595414A (en) | Heavy metal-polluted soil enterprise pollution source discrimination based on source remittance space variable reasoning | |
| CN110554013A (en) | method for realizing rapid identification and comparison by using fluorescence spectrum characteristic information | |
| CN110033133A (en) | Silt source tracing method in a kind of river and lake system | |
| Recanati et al. | Apatite (U-Th-Sm)/He date dispersion: First insights from machine learning algorithms | |
| Dayman et al. | Characterization of used nuclear fuel with multivariate analysis for process monitoring | |
| CN116187861A (en) | Isotope-based water quality traceability monitoring method and related device | |
| Frenzel | Making sense of mineral trace-element data–How to avoid common pitfalls in statistical analysis and interpretation | |
| CN108982402A (en) | The prediction technique and system of a kind of similar organic pollutant to environmental hazard | |
| Vovna et al. | Improving efficiency of information measurement system of coal mine air gas protection | |
| Lloyd et al. | Testing an improved index for analysing storm nutrient hysteresis | |
| CN109345074B (en) | Method for identifying and evaluating public acceptability influence factors of high-level waste geological disposal | |
| CN106778801A (en) | Method and apparatus for analyzing the classification of object of observation | |
| Cheng et al. | An Uncertainty Network Working Mechanism Analysis Method based on Local Interpretable Model-agnostic Explanations | |
| CN110807399A (en) | Single-category support vector machine-based collapse and slide hidden danger point detection method | |
| Luboschik et al. | Interactive visual exploration of simulator accuracy: A case study for stochastic simulation algorithms | |
| Raco et al. | GEOBASI: The geochemical Database of Tuscany Region (Italy) | |
| Moore et al. | Algorithm development for targeted isotopics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170510 |